Kathleen Folbigg's Pardon and Acquittal

How likely is it that the original judge got it so wrong when he found her guilty of murdering all four of her children?

During the last several days news has travelled around the world regarding the acquittal and pardon of Kathleen Folbigg. Folbigg was convicted and sentenced to 40 years imprisonment in 2003 for the suffocation deaths of all four of her infant children¹ born and deceased between 1989 and 1999. Her conviction came through the tragic misuse and mis-application of evidence, including misleading medical evidence². In many ways the original trial judgement in the Folbigg case mirrored issues observed in other similar ‘serial killer’ carer trials - such as those of Lucia De Berk, Sally Clark and more recently, Lucy Letby.

The forensic pathology evidence for the first three children had ruled their deaths as natural causes, and was initially indeterminate for the fourth. There was no physical evidence and no witness testimony that suggested culpability, or that Ms Folbigg was ever anything other than a distressed and grieving mother. This meant that the only tangible evidence prosecutors had to present at her original trial were some post-fact hand-written notes in diaries where Ms Folbigg in a clearly distraught state repeats what doctors, police and her husband had all said to her and, in essence, blames herself for the tragic deaths of her children. None of these notes could in any way be construed as an admission of guilt. However, the diary notes along with incriminating testimony by her own husband that impeached his prior police statement and the fact of her children’s deaths were all tentatively strung together using what has become known as ‘Meadows Law’ - the discredited claim of Roy Meadows that SIDS, then known as cot deaths but now referred to as SUDI³, were such a rare and unexplainable phenomena that it could be reasoned ‘one was a tragedy, two was suspicious, and three was murder unless proven to the contrary’. These few tenuous items were the sum total of what convicted Folbigg. To those who are aware of the cases of the other ‘serial killer’ carers mentioned in the previous paragraph, the similarities should now be readily apparent⁴.

In this article we will use a Bayesian network (BN) model to explore that original guilty verdict in light of the forensic pathologist’s reports, and culminating in an analysis showing strong support for the coroner’s acquital of Ms Folbigg this week on all charges.

The model

In this article we are going to consider three outcomes using a Bayesian network model that describes the key elements of the Folbigg case. This model will allow us to compute:

1. The likelihood that Folbigg murdered at least one of her children given the forensic pathology findings (forward reasoning);

2. The likelihood of the natural causes forensic findings given a guilty verdict that found Folgill murdered at least one of her children (backwards reasoning); and

3. The likelihood that the evidence supports the Coroner’s finding of natural causes deaths for all four children, and his acquittal of Kathleen Folbigg of having murdered any of them.

The BN model showing the prior probabilities for each node is shown below.

Finding it true that Folgill murdered at least one of her children should result in a guilty verdict, suggests it is more likely that she murdered some or all of the remaining children, and would justify the custodial sentence she received. Alternately, finding it false should result in an acquittal on all charges.

Given that until we observe the forensic pathologist’s autopsy findings either cause of death could be equally true, the priors for every one of the children’s nodes demonstrate our initial belief that each infant’s cause of death could equally be a natural occurence or murder. Each pathology report node’s prior state is weighted slightly towards a natural cause of death - outwardly reflecting a starting position of innocence until proven guilty, but inwardly ensuring we consider both the: (i) likely but low potential for and impact of error where the child’s cause of death is suspicious but the forensic pathologist has ruled it of natural causes; and (ii) slightly higher potential for and impact of error where the death is of natural causes but the forensic pathologist has ruled it suspicious.

The fourth and final pathology report node contains an additional node state - indeterminate. This is because, as you will see later, the original finding presented during the 2003 criminal trial for Folbigg’s fourth child was that her cause of death was unable to be determined.

The children’s nodes are interdependent. This represents the interrelationship between and impact of outcome on dependent variables, in that a finding of murder by Folgill for one child makes it more likely that the remaining children could have been murdered while conversly, a finding of natural causes for one child makes it slightly more likely for the remaining children’s deaths to also have a natural explanation.

Our model also considers the impact of the only other evidence presented at her original trial on the outcome of whether she murdered at least one of her children: (i) the notes in Folbigg’s diaries; and (ii) her husband’s incriminating testimony.

The First Three Children

On 20 February 1989 Kathleen Folbigg’s first child, Caleb, aged 19 days was established by ambulance officers to be in cardiac arrest and unable to be resuscitated. In the almost three weeks of his life, Caleb was clinically observed to have noisy breathing and bouts of respiratory distress, and was diagnosed with laryngomalacia - or floppy larynx. Caleb’s forensic autopsy report listed his cause of death as sudden infant death syndrome - or SIDS and noted the absence of any external signs of injury. Under Australian law a finding of SIDS is only open to the forensic pathologist if there are no suspicious circumstances surrounding the death. As such, SIDS is a diagnosis of exclusion based on the absence of any other possible cause of death, including murder.

Caleb was found to have died from natural causes.

On 13 February 1991 her second child, Patrick, died aged 8 months and 10 days at the Mater Hospital. During his short life Patrick had been clinically observed to have a large number of epileptic-type seizures that were described in court as approximately 15 fits diagnosed as intractable seizures, probably viral encephalitis or epilepsy. The forensic pathologist who performed the autopsy recorded the clinical diagnosis as encephalopathic disorder leading to intractable seizures. The underlying cause of the encephalopathy was not determined during the autopsy, but the final report stated that Patrick had an asystolic cardiac arrest at home leading to death.

Patrick was found to have died from natural causes.

On 30 August 1993 her third child, Sarah, died aged 10 months and 16 days when, after having provided almost 40 minutes of resuscitation, she became asystolic and paramedics withdrew care. Prof. John Hilton reported on the autopsy, finding that Sarah had died with focal pulmonary collapse, modest pulmonary congestion, and petechiae on the pleura, epicardia and in/on the thymus. He also noted that one section of the larynx showed a light mixed lymphatic inflammatory infiltrate deep to the respiratory epithelium.

Sarah was found to have died from natural causes.

We pause here to consider whether it would have been appropriate at this point to consider Folbigg a murder suspect. Using an abbreviated version of the model that incorporates nodes for only the first three children, we observe the only evidence available at this point - the forensic pathology findings that reported all three children died of natural causes.

We can see that with this limited evidence it was highly unlikely (94%) that Folbigg had murdered even one of her children, and therefore entirely appropriate not to have considered Folbigg a suspect at this point.

The Fourth Child

On 1 March 1999 Kathleen Folbigg’s fourth and final child, Laura, died aged 18 months and 22 days. Laura was clinically observed very early in her life to have some mild central apnoea that self-corrected. During June 1998 she was diagnosed with an upper respiratory tract infection and mild croup that proved to be recurrent and was diagnosed as a viral infection two months later in August 1988. She was also observed to have a viral macular skin rash and inflammed red throat and was prescribed the drug Phenergan. It should be noted that while Phenergan was likely used back then in order to both help Laura settle and possibly to rule out some sort of contact-based allergic reaction, we would not routinely use Phenergan in such small infants today due to the potential for fatal respiratory depression.

Dr Allan Cala authored the final forensic autopsy report, listing Laura’s cause of death as undetermined. He reported that she died with incidental findings of myocarditis and inflammation of the muscular walls of the heart. During the earlier 2019 enquiry Dr Cala gave evidence that on review of all materials in the case, he could not rule out myocarditis as the actual cause of Laura’s death.

Laura was initially found to have died from an indeterminate cause of death, but during the recent Coronial Inquest was found to have died from natural causes.

1. Guilty Verdict Given The Evidence

When we observe the forensic pathology findings on our BN model, we see that the Child nodes all strongly favour a natural, rather than murder, cause of death. We also see that it is highly likely (87%) that Folgill did not murder at least one of her children.

The model finds that, based solely on the forensic pathology evidence, the judge should acquit.

Even when we turn to consider the only other evidence against Folgill, the notes in her diaries and her former husband’s repudiation of his police statement and incriminating refreshed memory testimony led during trial, the model is still almost two-thirds (59.3%) in favour of the finding that she did not murder at least one of her children.

Yet, and in stark contrast, the original trial judge delivered a guilty verdict and sentenced Folgill to four decades in prison. This was a sentence that saw Folgill sequestered in protective custody for many years until, in 2021, she was brutally beaten by other inmates for her alleged crimes against children.

Therefore, the model shows that the evidence presented at trial did not naturally lead to the guilty verdict.

2. The Evidence Given A Guilty Verdict

Observation of the guilty verdict in Kathleen Folbigg’s original trial on the model results in a finding for each pathology report that is antithetical to the results reported for all of the children, but especially the first three.

The model identifies an almost two-thirds (64%) likelihood that the forensic pathologists would have had to have been suspicious of a non-natural cause of death in each case; sufficiently suspicious to support the findings of murder for at least one of the children. We can also observe that a suspicious death finding for even one of the children was sufficient to strongly support the inference that she murdered at least one of them…

And even if the first two had been ruled natural causes deaths…

This demonstrates that had Folbigg actually been guilty of murdering her children, it was increasingly more likely that evidence would have been identified during the autopsies suggesting questionable acts may have occured that led to the children’s deaths. However, this was not the case.

Therefore, the model shows that in order to arrive at a guilty verdict, evidence of suspicious forensic findings arising out of the autopsies was required.

3. The Case To Acquit

In the course of making his verdict the Coroner considered the updated testimony of Dr Cala and the recent published genetic science showing Folbigg’s daughters (children 3 and 4) had a genetic variation that causes cardiac arrhythmias which were likely instrumental in their deaths. He concluded that the fourth infant had also died of natural causes. Revising our model based on this updated information shows that the Folbigg is even less likely to have murdered even one of her children, with the probability that she did dropping to only 4.3%.

This supports the Coroner’s ultimate conclusion to acquit Kathleen Folbigg.

Finally, we turn to consider simply the case to acquit. In order to acquit it is necessary to find that Folbigg did not murder at least one of her children.

Backwards inference on the model of the Coroner’s decision to acquit renders results that more naturally accord with the original forensic pathology findings and genetic science. That recently published genetic science was instrumental in swaying the NSW Govenor to call for this inquest, and provides a simpler and more accurate explanation for the autopsy findings seen for both female children than the unsupported assertion from the original trial - that Kathleen Folbigg smothered them to death in their cribs. For each infant’s autopsy report we see support for an 80% or higher likelihood of a natural cause of death - including for the final and originally cause-indeterminate infant, Laura. As such, these results support the Coroner’s amended findings in Laura’s case - henceforth that her death now be recorded as one of natural causes.

Supported by the forensic pathology evidence for all four infants, while considering and discounting the improperly weighed evidence from the post-fact notes in Folbiggs diaries, the model supports the Coroner’s decision to acquit Kathleen Folbigg on all charges.

Conclusion

Without exception, criminal cases involving alleged harm against infants and children invoke strong emotions and unwavering determination to find and punish a guilty party. However, it can sometimes be that the person police, mainstream media and the public believe is the most likely purpetrator, is innocent. It can be that the alleged harm wasn’t harm at all - that there are other, more plausible but less understood medical reasons that explain why the child died. Like we have seen with other cases, such as those of Lucia De Berk and Sally Clark, this likely suspect as a result of police pertinacity and prosecutorial persuasiveness can be found guilty on little more than a few pieces of circumstantial evidence and the mere suggestion that they might have done something unthinkable.

Bayesian probabilistic models can and have previously been used to determine whether: (a) the observed evidence supports the verdict handed down; or (b) the type and strength of evidence required in order to support that same verdict. In this article we have used a Bayesian model to answer these questions for the Kathleen Folbigg case.

Our results were similar to the most recent Coronial findings in that we: (a) found that the evidence did not support the guilty verdict handed down in the original criminal trial; and that (b) evidence strongly suggestive of suspicious causes of death for at least one of the children would have been required. We also found that a judgement of acquittal not only more accurately aligns with the forensic evidence, but also supports the updated finding that Laura’s death was the result of natural causes - thus answering the final question that had remained open from the original trial and providing closure for the Folbigg family.

Similar models could be developed and applied in future on cases like that of Lucy Letby.

*** Addendum ***

Sydney lawyer Peter Gill has pointed out that I got two minor details wrong. First, even though from the transcripts I was reading it was apparent that a coroner had been involved in delivering some report during proceedings, the acquittal came from a judge (Judge Tom Bathurst KC) sitting in a ‘Special Inquiry’ not a ‘Coronial Inquiry’. The second item was that the Verdict and Sentencing Report I have did not mention that the original 20 year old guilty verdicts were delivered by jury deliberation, not by a judge alone (i.e. bench trial). These two items do not change the model or analysis, they would only change some of the wording or terms used in the narrative text of my article. For this reason I mention them here as an addendum rather than retroactively correcting them in the article.

***

1. Kathleen Folgill was found guilty of: (i) three counts of murder; (ii) one count of manslaughter; and (iii) one count of maliciously inflicting grievous bodily harm. She was sentenced to fourty years in prison with a non parole period of thirty years.

2. As laid out in legal academic Emma Cunliff’s 2011 text: Murder, Medicine and Motherhood.

3. SUDI: Sudden Unexplained Death in Infants

4. Sally Clark was convicted predominately on the basis of the flawed ‘Meadows Law’. Lucia De Berk and Lucy Letby were predominately convicted on the basis of their posession of nursing notes and, in Lucy’s case, the notes she wrote after her initial arrest that stated things like “they think I did it” and which are the product of a distraught and confused mind and not as the prosecution presented, an admission of culpability.

************************* ********************** *********************

I acknowledge the technical assistance of Professors Martin Neil and Norman Fenton during the development and refining of the model, analysis, and the writing of this article.